We have tested whether gap junctions form between nerve and muscle during their initial contact, before establishing the chemical synapse. Embryonic Xenopus stage 18-20 myotomes and neural tubes were permeabilized with DMSO to load appropriate reagents, dissociated, and cocultured. When myotomes, loaded with Lucifer yellow, were cocultured with unlabeled neural tube cells, 23% of the neurons contained dye after 24 hr. Affinity-purified gap junction antibodies loaded into myocytes or neurons reduced neuronal labeling significantly to 5%. [3H]uridine nucleotide transfer was observed in both directions between myocytes and neurons. Again gap junction antibodies substantially reduced recipient label. In all cases preimmune IgGs did not reduce transfer. When acetylcholine receptor clustering was examined in cultures containing gap junction antibodies, no difference in the number of neuronally induced AChR clusters was observed. This suggests that the cluster-inducing signal between nerve and muscle does not pass through gap junctions.